Performance over time

Tracking model performance as new versions and benchmarks arrive

A single leaderboard is a snapshot. As we re-test existing models on new versions and add new benchmarks beyond BR18, this page is where those results start showing real trends instead of one moment in time.

Tracking model performance as new versions and benchmarks arrive

Each point is one model’s accuracy on the day we tested it.

Hover a point for a quick look, or click it to keep the details open.

Benchmark

More benchmarks will appear here as we expand testing.

Explore the other views

Leaderboard

Accuracy by model

Point estimate with 90% confidence interval, sorted from highest to lowest accuracy.

Open
Calibration explorer

Does the model know what it doesn’t know?

For each self-reported confidence level, we plot the model’s actual accuracy.

Open